Financial market news portals are valuable sources of information as they hold great power over investors' decision-making processes. Due to the vast amount of text data produced by news portals, several studies have been conducted to comprehend the behavioral variations of texts and automate the categorization of short texts. However, extracting useful information that influences investors' decision-making process is not a trivial task, given that news portals use a heterogeneous and specific language for each content produced, making it challenging to generate a standard document format. This work proposes GOOSE, a solution for the cateGOrizatiOn of Short texts derived from multiple sources of information, to portray the financial market's current situation. To this end, GOOSE is based on Bidirectional Long Short-Term Memory (Bi-LSTM) and GloVe Embeddings to increase reliability in the short texts classification process. That way, GOOSE obtains data from news portals, which, once combined with a word embedding mechanism, are used as input for the Bi-LSTM to classify financial market news texts. The results obtained showed that GOOSE's efficiency in categorizing texts had an accuracy of 84% but also demonstrated the feasibility of its use in the extraction of information from financial market news portals.